/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * This code is issued under the GNU General Public License.
- * See the file COPYING in this distribution
+ * Copyright (c) 2002, 2003 Cluster File Systems, Inc.
*
- * Copyright (C) 1992, 1993, 1994, 1995
- * Remy Card (card@masi.ibp.fr)
- * Laboratoire MASI - Institut Blaise Pascal
- * Universite Pierre et Marie Curie (Paris VI)
+ * This file is part of Lustre, http://www.lustre.org.
*
- * from
+ * Lustre is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
*
- * linux/fs/ext2/namei.c
+ * Lustre is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with Lustre; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ * derived in small part from linux/fs/ext2/namei.c
*
* Copyright (C) 1991, 1992 Linus Torvalds
*
* David S. Miller (davem@caip.rutgers.edu), 1995
* Directory entry file type support and forward compatibility hooks
* for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998
- *
- * Changes for use in OBDFS
- * Copyright (c) 1999, Seagate Technology Inc.
- * Copyright (C) 2001, Cluster File Systems, Inc.
- * Rewritten based on recent ext2 page cache use.
- *
*/
#include <linux/fs.h>
#include <linux/obd_support.h>
#include <linux/lustre_lite.h>
#include <linux/lustre_dlm.h>
-#include <linux/obd_lov.h>
-
-extern struct address_space_operations ll_aops;
-
-/* from super.c */
-extern void ll_change_inode(struct inode *inode);
-extern int ll_setattr(struct dentry *de, struct iattr *attr);
-
-/* from dir.c */
-extern int ll_add_link (struct dentry *dentry, struct inode *inode);
-obd_id ll_inode_by_name(struct inode * dir, struct dentry *dentry, int *typ);
-int ext2_make_empty(struct inode *inode, struct inode *parent);
-struct ext2_dir_entry_2 * ext2_find_entry (struct inode * dir,
- struct dentry *dentry, struct page ** res_page);
-int ext2_delete_entry (struct ext2_dir_entry_2 * dir, struct page * page );
-int ext2_empty_dir (struct inode * inode);
-struct ext2_dir_entry_2 * ext2_dotdot (struct inode *dir, struct page **p);
-void ext2_set_link(struct inode *dir, struct ext2_dir_entry_2 *de,
- struct page *page, struct inode *inode);
+#include <linux/lustre_version.h>
+#include "llite_internal.h"
-/*
- * Couple of helper functions - make the code slightly cleaner.
- */
-static inline void ext2_inc_count(struct inode *inode)
-{
- inode->i_nlink++;
-}
+/* methods */
-/* postpone the disk update until the inode really goes away */
-static inline void ext2_dec_count(struct inode *inode)
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
+static int ll_test_inode(struct inode *inode, unsigned long ino, void *opaque)
+#else
+static int ll_test_inode(struct inode *inode, void *opaque)
+#endif
{
- inode->i_nlink--;
-}
+ static int last_ino, last_gen, last_count;
+ struct lustre_md *md = opaque;
-static inline int ext2_add_nondir(struct dentry *dentry, struct inode *inode)
-{
- int err;
- err = ll_add_link(dentry, inode);
- if (!err) {
- d_instantiate(dentry, inode);
+ if (!(md->body->valid & (OBD_MD_FLGENER | OBD_MD_FLID))) {
+ CERROR("MDS body missing inum or generation\n");
return 0;
}
- ext2_dec_count(inode);
- iput(inode);
- return err;
-}
-/* methods */
+ if (last_ino == md->body->ino && last_gen == md->body->generation &&
+ last_count < 500) {
+ last_count++;
+ } else {
+ if (last_count > 1)
+ CDEBUG(D_VFSTRACE, "compared %u/%u %u times\n",
+ last_ino, last_gen, last_count);
+ last_count = 0;
+ last_ino = md->body->ino;
+ last_gen = md->body->generation;
+ CDEBUG(D_VFSTRACE,
+ "comparing inode %p ino %lu/%u/%u to body %u/%u/%u\n",
+ inode, inode->i_ino, inode->i_generation,
+ ll_i2info(inode)->lli_mds,
+ md->body->ino, md->body->generation,
+ md->body->mds);
+ }
-#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
-static int ll_find_inode(struct inode *inode, unsigned long ino, void *opaque)
-#else
-static int ll_test_inode(struct inode *inode, void *opaque)
+#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
+ if (inode->i_ino != md->body->ino)
+ return 0;
#endif
-{
- struct ll_read_inode2_cookie *lic = opaque;
- struct mds_body *body = lic->lic_body;
+ if (inode->i_generation != md->body->generation)
+ return 0;
- if (inode->i_generation != lic->lic_body->generation)
+ if (ll_i2info(inode)->lli_mds != md->body->mds)
return 0;
/* Apply the attributes in 'opaque' to this inode */
- ll_update_inode(inode, body);
-
+ ll_update_inode(inode, md);
return 1;
}
RETURN(0);
}
+/* Get an inode by inode number (already instantiated by the intent lookup).
+ * Returns inode or NULL
+ */
#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
-extern int ll_read_inode2(struct inode *inode, void *opaque);
+int ll_set_inode(struct inode *inode, void *opaque)
+{
+ ll_read_inode2(inode, opaque);
+ return 0;
+}
+
struct inode *ll_iget(struct super_block *sb, ino_t hash,
- struct ll_read_inode2_cookie *lic)
+ struct lustre_md *md)
{
struct inode *inode;
- inode = iget5_locked(sb, hash, ll_test_inode, ll_read_inode2, lic);
-
- if (!inode)
- return ERR_PTR(-ENOMEM);
-
- if (inode->i_state & I_NEW) {
+ LASSERT(hash != 0);
+ inode = iget5_locked(sb, hash, ll_test_inode, ll_set_inode, md);
- unlock_new_inode(inode);
- }
+ if (inode) {
+ if (inode->i_state & I_NEW)
+ unlock_new_inode(inode);
+ CDEBUG(D_VFSTRACE, "inode: %lu/%u(%p)\n", inode->i_ino,
+ inode->i_generation, inode);
+ }
- // XXX Coda always fills inodes, should Lustre?
return inode;
}
#else
struct inode *ll_iget(struct super_block *sb, ino_t hash,
- struct ll_read_inode2_cookie *lic)
+ struct lustre_md *md)
{
struct inode *inode;
- inode = iget4(sb, hash, ll_find_inode, lic);
+ LASSERT(hash != 0);
+ inode = iget4(sb, hash, ll_test_inode, md);
+ if (inode)
+ CDEBUG(D_VFSTRACE, "inode: %lu/%u(%p)\n", inode->i_ino,
+ inode->i_generation, inode);
return inode;
}
#endif
-static int ll_intent_to_lock_mode(struct lookup_intent *it)
+int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
+ void *data, int flag)
{
- /* CREAT needs to be tested before open (both could be set) */
- if ((it->it_op & (IT_CREAT | IT_MKDIR | IT_SETATTR | IT_MKNOD))) {
- return LCK_PW;
- } else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_UNLINK |
- IT_RMDIR | IT_RENAME | IT_RENAME2 | IT_READLINK|
- IT_LINK | IT_LINK2 | IT_LOOKUP | IT_SYMLINK)) {
- return LCK_PW;
- }
-
- LBUG();
- RETURN(-EINVAL);
-}
-
-#define LL_LOOKUP_POSITIVE 1
-#define LL_LOOKUP_NEGATIVE 2
-
-int ll_intent_lock(struct inode *parent, struct dentry **de,
- struct lookup_intent *it,
- intent_finish_cb intent_finish)
-{
- struct dentry *dentry = *de;
- struct ll_sb_info *sbi = ll_i2sbi(parent);
+ int rc;
struct lustre_handle lockh;
- struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
- struct ptlrpc_request *request = NULL;
- char *tgt = NULL;
- int rc, lock_mode, tgtlen = 0, offset, flag = LL_LOOKUP_POSITIVE;
- obd_id ino = 0;
-
ENTRY;
- if (it == NULL)
- it = &lookup_it;
-
- CDEBUG(D_INFO, "name: %*s, intent: %s\n", dentry->d_name.len,
- dentry->d_name.name, ldlm_it2str(it->it_op));
-
- if (dentry->d_name.len > EXT2_NAME_LEN)
- RETURN(-ENAMETOOLONG);
-
- lock_mode = ll_intent_to_lock_mode(it);
- if (it->it_op & IT_SYMLINK) {
- tgt = it->it_data;
- tgtlen = strlen(tgt);
- it->it_data = NULL;
- }
-
- rc = mdc_enqueue(&sbi->ll_mdc_conn, LDLM_PLAIN, it, lock_mode, parent,
- dentry, &lockh, tgt, tgtlen, parent, sizeof(*parent));
- if (rc < 0)
- RETURN(rc);
- memcpy(it->it_lock_handle, &lockh, sizeof(lockh));
-
- request = (struct ptlrpc_request *)it->it_data;
- if (it->it_disposition) {
- struct mds_body *mds_body;
- int mode, symlen = 0;
- obd_flag valid;
-
- /* it_disposition == 1 indicates that the server performed the
- * intent on our behalf. This long block is all about fixing up
- * the local state so that it is correct as of the moment
- * _before_ the operation was applied; that way, the VFS will
- * think that everything is normal and call Lustre's regular
- * FS function.
- *
- * If we're performing a creation, that means that unless the
- * creation failed with EEXIST, we should fake up a negative
- * dentry. Likewise for the target of a hard link.
- *
- * For everything else, we want to lookup to succeed. */
-
- /* One additional note: we add an extra reference to the request
- * because we need to keep it around until ll_create gets
- * called. For anything else which results in
- * LL_LOOKUP_POSITIVE, we can do the iget() immediately with the
- * contents of the reply (in the intent_finish callback). In
- * the create case, however, we need to wait until
- * ll_create_node to do the iget() or the VFS will abort with
- * -EEXISTS. */
-
- offset = 1;
- mds_body = lustre_msg_buf(request->rq_repmsg, offset);
- ino = mds_body->fid1.id;
- mode = mds_body->mode;
- if (it->it_op & (IT_CREAT | IT_MKDIR | IT_SYMLINK | IT_MKNOD)) {
- mdc_store_create_replay_data(request, parent->i_sb);
- /* For create ops, we want the lookup to be negative,
- * unless the create failed in a way that indicates
- * that the file is already there */
- if (it->it_status != -EEXIST) {
- atomic_inc(&request->rq_refcount);
- GOTO(out, flag = LL_LOOKUP_NEGATIVE);
- }
- /* Fall through to update attibutes. */
- } else if (it->it_op & (IT_GETATTR | IT_SETATTR | IT_LOOKUP |
- IT_READLINK)) {
- /* For check ops, we want the lookup to succeed */
- it->it_data = NULL;
- if (it->it_status)
- GOTO(out, flag = LL_LOOKUP_NEGATIVE);
- /* Fall through to update attibutes. */
- } else if (it->it_op & (IT_RENAME | IT_LINK)) {
- /* For rename, we want the source lookup to succeed */
- if (it->it_status) {
- it->it_data = NULL;
- GOTO(drop_req, rc = it->it_status);
- }
- it->it_data = dentry;
- /* Fall through to update attibutes. */
- } else if (it->it_op & (IT_UNLINK | IT_RMDIR)) {
- /* For remove ops, we want the lookup to succeed unless
- * the file truly doesn't exist */
- it->it_data = NULL;
- if (it->it_status == -ENOENT)
- GOTO(out, flag = LL_LOOKUP_NEGATIVE);
- /* No point in updating attributes that we're about to
- * unlink. -phil */
- GOTO(out, flag = LL_LOOKUP_POSITIVE);
- } else if (it->it_op == IT_OPEN) {
- it->it_data = NULL;
- if (it->it_status && it->it_status != -EEXIST)
- GOTO(out, flag = LL_LOOKUP_NEGATIVE);
- /* Fall through to update attibutes. */
- } else if (it->it_op & (IT_RENAME2 | IT_LINK2)) {
- it->it_data = NULL;
- /* This means the target lookup is negative */
- if (mds_body->valid == 0)
- GOTO(out, flag = LL_LOOKUP_NEGATIVE);
- /* XXX bug 289: should we maybe fall through here? -p */
- GOTO(out, flag = LL_LOOKUP_POSITIVE);
+ switch (flag) {
+ case LDLM_CB_BLOCKING:
+ ldlm_lock2handle(lock, &lockh);
+ rc = ldlm_cli_cancel(&lockh);
+ if (rc < 0) {
+ CDEBUG(D_INODE, "ldlm_cli_cancel: %d\n", rc);
+ RETURN(rc);
}
+ break;
+ case LDLM_CB_CANCELING: {
+ struct inode *inode = ll_inode_from_lock(lock);
+ __u64 bits = lock->l_policy_data.l_inodebits.bits;
- /* Do a getattr now that we have the lock */
- valid = OBD_MD_FLNOTOBD | OBD_MD_FLEASIZE;
- if (it->it_op == IT_READLINK) {
- valid |= OBD_MD_LINKNAME;
- symlen = mds_body->size;
- }
- ptlrpc_req_finished(request);
- request = NULL;
- rc = mdc_getattr(&sbi->ll_mdc_conn, ino, mode,
- valid, symlen, &request);
- if (rc) {
- CERROR("failure %d inode "LPX64"\n", rc, ino);
- GOTO(drop_req, rc = -abs(rc));
- }
- offset = 0;
- } else {
- struct ll_inode_info *lli = ll_i2info(parent);
- int mode;
+ /* For lookup locks: Invalidate all dentries associated with
+ this inode, for UPDATE locks - invalidate directory pages */
+ if (inode == NULL)
+ break;
- /* it_disposition == 0 indicates that it just did a simple lock
- * request, for which we are very thankful. move along with
- * the local lookup then. */
+ if (bits & MDS_INODELOCK_UPDATE)
+ clear_bit(LLI_F_HAVE_MDS_SIZE_LOCK,
+ &(ll_i2info(inode)->lli_flags));
- memcpy(&lli->lli_intent_lock_handle, &lockh, sizeof(lockh));
- offset = 0;
- ino = ll_inode_by_name(parent, dentry, &mode);
- if (!ino) {
- CERROR("inode %*s not found by name\n",
- dentry->d_name.len, dentry->d_name.name);
- GOTO(drop_lock, rc = -ENOENT);
+ if (lock->l_resource->lr_name.name[0] != inode->i_ino ||
+ lock->l_resource->lr_name.name[1] != inode->i_generation) {
+ LDLM_ERROR(lock, "data mismatch with ino %lu/%u",
+ inode->i_ino, inode->i_generation);
}
- rc = mdc_getattr(&sbi->ll_mdc_conn, ino, mode,
- OBD_MD_FLNOTOBD|OBD_MD_FLEASIZE, 0, &request);
- if (rc) {
- CERROR("failure %d inode "LPX64"\n", rc, ino);
- GOTO(drop_req, rc = -abs(rc));
+ /* If lookup lock is cancelled, we just drop the dentry and
+ this will cause us to reget data from MDS when we'd want to
+ access this dentry/inode again. If this is lock on
+ other parts of inode that is cancelled, we do not need to do
+ much (but need to discard data from readdir, if any), since
+ abscence of lock will cause ll_revalidate_it (called from
+ stat() and similar functions) to renew the data anyway */
+ if (S_ISDIR(inode->i_mode) &&
+ (bits & MDS_INODELOCK_UPDATE)) {
+ CDEBUG(D_INODE, "invalidating inode %lu\n",
+ inode->i_ino);
+
+ truncate_inode_pages(inode->i_mapping, 0);
}
- }
- EXIT;
- out:
- if (intent_finish != NULL) {
- rc = intent_finish(flag, request, de, it, offset, ino);
- dentry = *de; /* intent_finish may change *de */
- } else {
- ptlrpc_req_finished(request);
+ if (inode->i_sb->s_root &&
+ inode != inode->i_sb->s_root->d_inode &&
+ (bits & MDS_INODELOCK_LOOKUP))
+ ll_unhash_aliases(inode);
+ iput(inode);
+ break;
+ }
+ default:
+ LBUG();
}
- if (it->it_op == IT_LOOKUP || rc < 0)
- ll_intent_release(dentry, it);
-
- return rc;
+ RETURN(0);
+}
- drop_req:
- ptlrpc_free_req(request);
- drop_lock:
-#warning FIXME: must release lock here
- return rc;
+int ll_mdc_cancel_unused(struct lustre_handle *conn, struct inode *inode,
+ int flags, void *opaque)
+{
+ struct ldlm_res_id res_id =
+ { .name = {inode->i_ino, inode->i_generation} };
+ struct obd_device *obddev = class_conn2obd(conn);
+ ENTRY;
+
+ RETURN(ldlm_cli_cancel_unused(obddev->obd_namespace, &res_id, flags,
+ opaque));
}
/* Search "inode"'s alias list for a dentry that has the same name and parent as
* de. If found, return it. If not found, return de. */
-static struct dentry *ll_find_alias(struct inode *inode, struct dentry *de)
+struct dentry *ll_find_alias(struct inode *inode, struct dentry *de)
{
- struct list_head *tmp;
+ struct list_head *tmp;
- spin_lock(&dcache_lock);
+ spin_lock(&dcache_lock);
list_for_each(tmp, &inode->i_dentry) {
- struct dentry *dentry = list_entry(tmp, struct dentry, d_alias);
+ struct dentry *dentry = list_entry(tmp, struct dentry, d_alias);
/* We are called here with 'de' already on the aliases list. */
- if (dentry == de)
+ if (dentry == de) {
+ CERROR("whoops\n");
continue;
+ }
if (dentry->d_parent != de->d_parent)
continue;
de->d_name.len) != 0)
continue;
+ if (!list_empty(&dentry->d_lru))
+ list_del_init(&dentry->d_lru);
+
+ hlist_del_init(&dentry->d_hash);
+ __d_rehash(dentry, 0); /* avoid taking dcache_lock inside */
spin_unlock(&dcache_lock);
- d_rehash(dentry);
- return dget(dentry);
- }
- spin_unlock(&dcache_lock);
+ atomic_inc(&dentry->d_count);
+ iput(inode);
+ dentry->d_flags &= ~DCACHE_LUSTRE_INVALID;
+ CDEBUG(D_DENTRY, "alias dentry %*s (%p) parent %p inode %p "
+ "refc %d\n", de->d_name.len, de->d_name.name, de,
+ de->d_parent, de->d_inode, atomic_read(&de->d_count));
+ return dentry;
+ }
+
+ spin_unlock(&dcache_lock);
+
return de;
}
-static int
-lookup2_finish(int flag, struct ptlrpc_request *request, struct dentry **de,
- struct lookup_intent *it, int offset, obd_id ino)
+static int lookup_it_finish(struct ptlrpc_request *request, int offset,
+ struct lookup_intent *it, void *data)
{
+ struct it_cb_data *icbd = data;
+ struct dentry **de = icbd->icbd_childp;
+ struct inode *parent = icbd->icbd_parent;
+ struct ll_sb_info *sbi = ll_i2sbi(parent);
struct dentry *dentry = *de, *saved = *de;
struct inode *inode = NULL;
- struct ll_read_inode2_cookie lic;
+ int rc;
- if (flag == LL_LOOKUP_POSITIVE) {
+ /* NB 1 request reference will be taken away by ll_intent_lock()
+ * when I return */
+ if (!it_disposition(it, DISP_LOOKUP_NEG)) {
ENTRY;
- lic.lic_body = lustre_msg_buf(request->rq_repmsg, offset);
-
- if (S_ISREG(lic.lic_body->mode) &&
- lic.lic_body->valid & OBD_MD_FLEASIZE) {
- LASSERT(request->rq_repmsg->bufcount > offset);
- lic.lic_lmm = lustre_msg_buf(request->rq_repmsg,
- offset + 1);
- } else {
- lic.lic_lmm = NULL;
- }
- /* No rpc's happen during iget4, -ENOMEM's are possible */
- LASSERT(ino != 0);
- inode = ll_iget(dentry->d_sb, ino, &lic);
- if (!inode) {
- /* XXX make sure that request is freed in this case;
- * I think it is, but double-check refcounts. -phil */
- RETURN(-ENOMEM);
- }
+ rc = ll_prep_inode(sbi->ll_osc_exp, sbi->ll_mdc_exp,
+ &inode, request, offset, dentry->d_sb);
+ if (rc)
+ RETURN(rc);
- dentry = *de = ll_find_alias(inode, dentry);
+ CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
+ inode, inode->i_ino, inode->i_generation);
+ mdc_set_lock_data(NULL, &it->d.lustre.it_lock_handle, inode);
+
+ /* If this is a stat, get the authoritative file size */
+ if (it->it_op == IT_GETATTR && S_ISREG(inode->i_mode) &&
+ ll_i2info(inode)->lli_smd != NULL) {
+ struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
+ struct ost_lvb lvb;
+ ldlm_error_t rc;
+
+ LASSERT(lsm->lsm_object_id != 0);
- /* We asked for a lock on the directory, and may have been
- * granted a lock on the inode. Just in case, fixup the data
- * pointer. */
- ldlm_lock_set_data((struct lustre_handle *)it->it_lock_handle,
- inode, sizeof(*inode));
+ /* bug 2334: drop MDS lock before acquiring OST lock */
+ ll_intent_drop_lock(it);
- EXIT;
+ rc = ll_glimpse_size(inode, &lvb);
+ if (rc) {
+ iput(inode);
+ RETURN(rc);
+ }
+ inode->i_size = lvb.lvb_size;
+ }
+
+ dentry = *de = ll_find_alias(inode, dentry);
} else {
ENTRY;
}
- ptlrpc_req_finished(request);
-
dentry->d_op = &ll_d_ops;
- if (ll_d2d(dentry) == NULL) {
- ll_set_dd(dentry);
- }
+ ll_set_dd(dentry);
- if (dentry == saved)
+ if (dentry == saved) {
d_add(dentry, inode);
-
- if (it->it_status == 0) {
- LL_SAVE_INTENT(dentry, it);
- } else {
- dentry->d_it = NULL;
- CDEBUG(D_DENTRY,
- "D_IT dentry %p fsdata %p intent: %s status %d\n",
- dentry, ll_d2d(dentry), ldlm_it2str(it->it_op),
- it->it_status);
}
RETURN(0);
}
-static struct dentry *ll_lookup2(struct inode *parent, struct dentry *dentry,
- struct lookup_intent *it)
+
+static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry,
+ struct nameidata *nd,
+ struct lookup_intent *it, int flags)
{
- struct dentry *save = dentry;
+ struct dentry *save = dentry, *retval;
+ struct ll_fid pfid;
+ struct ll_uctxt ctxt;
+ struct it_cb_data icbd;
+ struct ptlrpc_request *req = NULL;
+ struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
int rc;
+ ENTRY;
- rc = ll_intent_lock(parent, &dentry, it, lookup2_finish);
- if (rc < 0) {
- CERROR("ll_intent_lock: %d\n", rc);
- return ERR_PTR(rc);
- }
+ if (dentry->d_name.len > EXT3_NAME_LEN)
+ RETURN(ERR_PTR(-ENAMETOOLONG));
- if (dentry == save)
- return NULL;
- else
- return dentry;
-}
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p),intent=%s\n",
+ dentry->d_name.name, parent->i_ino, parent->i_generation,
+ parent, LL_IT2STR(it));
-static struct inode *ll_create_node(struct inode *dir, const char *name,
- int namelen, const char *tgt, int tgtlen,
- int mode, __u64 extra,
- struct lookup_intent *it,
- struct lov_stripe_md *lsm)
-{
- struct inode *inode;
- struct ptlrpc_request *request = NULL;
- struct mds_body *body;
- time_t time = CURRENT_TIME;
- struct ll_sb_info *sbi = ll_i2sbi(dir);
- struct ll_read_inode2_cookie lic;
- struct lov_mds_md *lmm = NULL;
- ENTRY;
+ if (d_mountpoint(dentry))
+ CERROR("Tell Peter, lookup on mtpt, it %s\n", LL_IT2STR(it));
- if (it && it->it_disposition) {
- int rc = it->it_status;
- if (rc) {
- CERROR("error creating MDS inode for %*s: rc = %d\n",
- namelen, name, rc);
- RETURN(ERR_PTR(rc));
- }
- ll_invalidate_inode_pages(dir);
- request = it->it_data;
- body = lustre_msg_buf(request->rq_repmsg, 1);
- lic.lic_lmm = NULL;
- } else {
- int gid = current->fsgid;
- int rc;
-
- if (lsm) {
- OBD_ALLOC(lmm, lsm->lsm_mds_easize);
- if (!lmm)
- RETURN(ERR_PTR(-ENOMEM));
- lov_packmd(lmm, lsm);
- lic.lic_lmm = lmm;
- } else
- lic.lic_lmm = NULL;
-
- if (dir->i_mode & S_ISGID) {
- gid = dir->i_gid;
- if (S_ISDIR(mode))
- mode |= S_ISGID;
- }
+ if (nd != NULL)
+ nd->mnt->mnt_last_used = jiffies;
- rc = mdc_create(&sbi->ll_mdc_conn, dir, name, namelen, tgt,
- tgtlen, mode, current->fsuid, gid,
- time, extra, lsm, &request);
- if (rc) {
- inode = ERR_PTR(rc);
- GOTO(out, rc);
- }
- body = lustre_msg_buf(request->rq_repmsg, 0);
- }
+ ll_frob_intent(&it, &lookup_it);
- lic.lic_body = body;
+ icbd.icbd_childp = &dentry;
+ icbd.icbd_parent = parent;
+ ll_inode2fid(&pfid, parent);
+ ll_i2uctxt(&ctxt, parent, NULL);
- LASSERT(body->ino != 0);
- inode = ll_iget(dir->i_sb, body->ino, &lic);
- if (IS_ERR(inode)) {
- int rc = PTR_ERR(inode);
- CERROR("new_inode -fatal: rc %d\n", rc);
- LBUG();
- GOTO(out, rc);
- }
+ rc = md_intent_lock(ll_i2mdcexp(parent), &ctxt, &pfid,
+ dentry->d_name.name, dentry->d_name.len, NULL, 0,
+ NULL, it, flags, &req, ll_mdc_blocking_ast);
+ if (rc < 0)
+ GOTO(out, retval = ERR_PTR(rc));
- if (!list_empty(&inode->i_dentry)) {
- CERROR("new_inode -fatal: inode %d, ct %d lnk %d\n",
- body->ino, atomic_read(&inode->i_count),
- inode->i_nlink);
- iput(inode);
- LBUG();
- inode = ERR_PTR(-EIO);
- GOTO(out, -EIO);
+ rc = lookup_it_finish(req, 1, it, &icbd);
+ if (rc != 0) {
+ ll_intent_release(it);
+ GOTO(out, retval = ERR_PTR(rc));
}
- if (it && it->it_disposition) {
- /* We asked for a lock on the directory, but were
- * granted a lock on the inode. Since we finally have
- * an inode pointer, stuff it in the lock. */
- ldlm_lock_set_data((struct lustre_handle *)it->it_lock_handle,
- inode, sizeof(*inode));
- }
+ ll_lookup_finish_locks(it, dentry);
- EXIT;
+ if (nd &&
+ dentry->d_inode != NULL && dentry->d_inode->i_mode & S_ISUID &&
+ S_ISDIR(dentry->d_inode->i_mode) &&
+ (flags & LOOKUP_CONTINUE || (it->it_op & (IT_CHDIR | IT_OPEN))))
+ ll_dir_process_mount_object(dentry, nd->mnt);
+
+ if (dentry == save)
+ GOTO(out, retval = NULL);
+ else
+ GOTO(out, retval = dentry);
out:
- if (lsm && lmm)
- OBD_FREE(lmm, lsm->lsm_mds_easize);
- ptlrpc_req_finished(request);
- return inode;
+ if (req)
+ ptlrpc_req_finished(req);
+ if (dentry->d_inode)
+ CDEBUG(D_INODE, "lookup 0x%p in %lu/%lu: %*s -> %lu/%lu\n",
+ dentry,
+ (unsigned long) parent->i_ino,
+ (unsigned long) parent->i_generation,
+ dentry->d_name.len, dentry->d_name.name,
+ (unsigned long) dentry->d_inode->i_ino,
+ (unsigned long) dentry->d_inode->i_generation);
+ else
+ CDEBUG(D_INODE, "lookup 0x%p in %lu/%lu: %*s -> ??\n",
+ dentry,
+ (unsigned long) parent->i_ino,
+ (unsigned long) parent->i_generation,
+ dentry->d_name.len, dentry->d_name.name);
+ return retval;
}
-static int ll_mdc_unlink(struct inode *dir, struct inode *child, __u32 mode,
- const char *name, int len)
+#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
+static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry,
+ struct nameidata *nd)
{
- struct ptlrpc_request *request = NULL;
- struct ll_sb_info *sbi = ll_i2sbi(dir);
- int err;
-
+ struct dentry *de;
ENTRY;
- err = mdc_unlink(&sbi->ll_mdc_conn, dir, child, mode, name, len,
- &request);
- ptlrpc_req_finished(request);
+ if (nd && nd->flags & LOOKUP_LAST && !(nd->flags & LOOKUP_LINK_NOTLAST))
+ de = ll_lookup_it(parent, dentry, nd, &nd->intent, nd->flags);
+ else
+ de = ll_lookup_it(parent, dentry, nd, NULL, 0);
- RETURN(err);
+ RETURN(de);
}
+#endif
-int ll_mdc_link(struct dentry *src, struct inode *dir,
- const char *name, int len)
+/* We depend on "mode" being set with the proper file type/umask by now */
+static struct inode *ll_create_node(struct inode *dir, const char *name,
+ int namelen, const void *data, int datalen,
+ int mode, __u64 extra,
+ struct lookup_intent *it)
{
+ struct inode *inode = NULL;
struct ptlrpc_request *request = NULL;
- int err;
struct ll_sb_info *sbi = ll_i2sbi(dir);
-
+ int rc;
ENTRY;
- err = mdc_link(&sbi->ll_mdc_conn, src, dir, name, len, &request);
- ptlrpc_req_finished(request);
-
- RETURN(err);
-}
+ LASSERT(it && it->d.lustre.it_disposition);
-int ll_mdc_rename(struct inode *src, struct inode *tgt,
- struct dentry *old, struct dentry *new)
-{
- struct ptlrpc_request *request = NULL;
- struct ll_sb_info *sbi = ll_i2sbi(src);
- int err;
+ request = it->d.lustre.it_data;
+ rc = ll_prep_inode(sbi->ll_osc_exp, sbi->ll_mdc_exp,
+ &inode, request, 1, dir->i_sb);
+ if (rc)
+ GOTO(out, inode = ERR_PTR(rc));
- ENTRY;
+ LASSERT(list_empty(&inode->i_dentry));
- err = mdc_rename(&sbi->ll_mdc_conn, src, tgt,
- old->d_name.name, old->d_name.len,
- new->d_name.name, new->d_name.len, &request);
+ /* We asked for a lock on the directory, but were granted a
+ * lock on the inode. Since we finally have an inode pointer,
+ * stuff it in the lock. */
+ CDEBUG(D_DLMTRACE, "setting l_ast_data to inode %p (%lu/%u)\n",
+ inode, inode->i_ino, inode->i_generation);
+ mdc_set_lock_data(NULL, &it->d.lustre.it_lock_handle, inode);
+ EXIT;
+ out:
ptlrpc_req_finished(request);
-
- RETURN(err);
+ return inode;
}
/*
* By the time this is called, we already have created the directory cache
* entry for the new file, but it is so far negative - it has no inode.
+ *
* We defer creating the OBD object(s) until open, to keep the intent and
* non-intent code paths similar, and also because we do not have the MDS
* inode number before calling ll_create_node() (which is needed for LOV),
* so we would need to do yet another RPC to the MDS to store the LOV EA
- * data on the MDS.
+ * data on the MDS. If needed, we would pass the PACKED lmm as data and
+ * lmm_size in datalen (the MDS still has code which will handle that).
*
* If the create succeeds, we fill in the inode information
* with d_instantiate().
*/
-static int ll_create(struct inode *dir, struct dentry *dentry, int mode)
+static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode,
+ struct lookup_intent *it)
{
- struct lookup_intent *it;
struct inode *inode;
+ struct ptlrpc_request *request = it->d.lustre.it_data;
+ struct obd_export *mdc_exp = ll_i2mdcexp(dir);
int rc = 0;
ENTRY;
- LL_GET_INTENT(dentry, it);
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p),intent=%s\n",
+ dentry->d_name.name, dir->i_ino, dir->i_generation, dir,
+ LL_IT2STR(it));
- inode = ll_create_node(dir, dentry->d_name.name, dentry->d_name.len,
- NULL, 0, mode, 0, it, NULL);
+ rc = it_open_error(DISP_OPEN_CREATE, it);
+ if (rc)
+ RETURN(rc);
- if (IS_ERR(inode))
+ mdc_store_inode_generation(mdc_exp, request, 2, 1);
+ inode = ll_create_node(dir, dentry->d_name.name, dentry->d_name.len,
+ NULL, 0, mode, 0, it);
+ if (IS_ERR(inode)) {
RETURN(PTR_ERR(inode));
-
- if (it->it_disposition) {
- struct ll_inode_info *lli = ll_i2info(inode);
- memcpy(&lli->lli_intent_lock_handle, it->it_lock_handle,
- sizeof(lli->lli_intent_lock_handle));
- d_instantiate(dentry, inode);
- } else {
- /* no directory data updates when intents rule */
- rc = ext2_add_nondir(dentry, inode);
}
- RETURN(rc);
+ d_instantiate(dentry, inode);
+ RETURN(0);
}
-static int ll_mknod(struct inode *dir, struct dentry *dentry, int mode,
- int rdev)
+#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
+static int ll_create_nd(struct inode *dir, struct dentry *dentry, int mode, struct nameidata *nd)
{
- struct lookup_intent *it;
- struct inode *inode;
- int rc = 0;
-
- LL_GET_INTENT(dentry, it);
-
- inode = ll_create_node(dir, dentry->d_name.name, dentry->d_name.len,
- NULL, 0, mode, rdev, it, NULL);
-
- if (IS_ERR(inode))
- RETURN(PTR_ERR(inode));
+ return ll_create_it(dir, dentry, mode, &nd->intent);
+}
+#endif
- /* no directory data updates when intents rule */
- if (it && it->it_disposition)
- d_instantiate(dentry, inode);
- else
- rc = ext2_add_nondir(dentry, inode);
+static int ll_mknod_raw(struct nameidata *nd, int mode, dev_t rdev)
+{
+ struct ptlrpc_request *request = NULL;
+ struct inode *dir = nd->dentry->d_inode;
+ const char *name = nd->last.name;
+ int len = nd->last.len;
+ struct ll_sb_info *sbi = ll_i2sbi(dir);
+ struct mdc_op_data op_data;
+ int err = -EMLINK;
+ ENTRY;
- return rc;
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
+ name, dir->i_ino, dir->i_generation, dir);
+
+ if (dir->i_nlink >= EXT3_LINK_MAX)
+ RETURN(err);
+
+ mode &= ~current->fs->umask;
+
+ switch (mode & S_IFMT) {
+ case 0:
+ case S_IFREG:
+ mode |= S_IFREG; /* for mode = 0 case, fallthrough */
+ case S_IFCHR:
+ case S_IFBLK:
+ case S_IFIFO:
+ case S_IFSOCK:
+ ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
+ err = md_create(sbi->ll_mdc_exp, &op_data, NULL, 0, mode,
+ current->fsuid, current->fsgid, rdev, &request);
+ ptlrpc_req_finished(request);
+ break;
+ case S_IFDIR:
+ err = -EPERM;
+ break;
+ default:
+ err = -EINVAL;
+ }
+ RETURN(err);
}
-static int ll_symlink(struct inode *dir, struct dentry *dentry,
- const char *symname)
+static int ll_mknod(struct inode *dir, struct dentry *child, int mode,
+ ll_dev_t rdev)
{
- struct lookup_intent *it;
- unsigned l = strlen(symname);
- struct inode *inode;
- struct ll_inode_info *lli;
- int err = 0;
+ struct ptlrpc_request *request = NULL;
+ struct inode *inode = NULL;
+ const char *name = child->d_name.name;
+ int len = child->d_name.len;
+ struct ll_sb_info *sbi = ll_i2sbi(dir);
+ struct mdc_op_data op_data;
+ int err = -EMLINK;
ENTRY;
- LL_GET_INTENT(dentry, it);
-
- inode = ll_create_node(dir, dentry->d_name.name, dentry->d_name.len,
- symname, l, S_IFLNK | S_IRWXUGO, 0, it, NULL);
- if (IS_ERR(inode))
- RETURN(PTR_ERR(inode));
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
+ name, dir->i_ino, dir->i_generation, dir);
+
+ if (dir->i_nlink >= EXT3_LINK_MAX)
+ RETURN(err);
+
+ mode &= ~current->fs->umask;
+
+ switch (mode & S_IFMT) {
+ case 0:
+ case S_IFREG:
+ mode |= S_IFREG; /* for mode = 0 case, fallthrough */
+ case S_IFCHR:
+ case S_IFBLK:
+ case S_IFIFO:
+ case S_IFSOCK:
+ ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
+ err = md_create(sbi->ll_mdc_exp, &op_data, NULL, 0, mode,
+ current->fsuid, current->fsgid, rdev, &request);
+ err = ll_prep_inode(sbi->ll_osc_exp, sbi->ll_mdc_exp,
+ &inode, request, 0, child->d_sb);
+ if (err)
+ GOTO(out_err, err);
+ break;
+ case S_IFDIR:
+ RETURN(-EPERM);
+ break;
+ default:
+ RETURN(-EINVAL);
+ }
- lli = ll_i2info(inode);
+ d_instantiate(child, inode);
+ out_err:
+ ptlrpc_req_finished(request);
+ RETURN(err);
+}
- OBD_ALLOC(lli->lli_symlink_name, l + 1);
- /* this _could_ be a non-fatal error, since the symlink is already
- * stored on the MDS by this point, and we can re-get it in readlink.
- */
- if (!lli->lli_symlink_name)
- RETURN(-ENOMEM);
+static int ll_symlink_raw(struct nameidata *nd, const char *tgt)
+{
+ struct inode *dir = nd->dentry->d_inode;
+ const char *name = nd->last.name;
+ int len = nd->last.len;
+ struct ptlrpc_request *request = NULL;
+ struct ll_sb_info *sbi = ll_i2sbi(dir);
+ struct mdc_op_data op_data;
+ int err = -EMLINK;
+ ENTRY;
- memcpy(lli->lli_symlink_name, symname, l + 1);
- inode->i_size = l;
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p),target=%s\n",
+ name, dir->i_ino, dir->i_generation, dir, tgt);
- /* no directory data updates when intents rule */
- if (it && it->it_disposition)
- d_instantiate(dentry, inode);
- else
- err = ext2_add_nondir(dentry, inode);
+ if (dir->i_nlink >= EXT3_LINK_MAX)
+ RETURN(err);
+ ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
+ err = md_create(sbi->ll_mdc_exp, &op_data,
+ tgt, strlen(tgt) + 1, S_IFLNK | S_IRWXUGO,
+ current->fsuid, current->fsgid, 0, &request);
+ ptlrpc_req_finished(request);
RETURN(err);
}
-static int ll_link(struct dentry *old_dentry, struct inode * dir,
- struct dentry *dentry)
+static int ll_link_raw(struct nameidata *srcnd, struct nameidata *tgtnd)
{
- struct lookup_intent *it;
- struct inode *inode = old_dentry->d_inode;
- int rc;
-
- LL_GET_INTENT(dentry, it);
+ struct inode *src = srcnd->dentry->d_inode;
+ struct inode *dir = tgtnd->dentry->d_inode;
+ const char *name = tgtnd->last.name;
+ int len = tgtnd->last.len;
+ struct ptlrpc_request *request = NULL;
+ struct mdc_op_data op_data;
+ int err;
+ struct ll_sb_info *sbi = ll_i2sbi(dir);
- if (it && it->it_disposition) {
- if (it->it_status)
- RETURN(it->it_status);
- inode->i_ctime = CURRENT_TIME;
- ext2_inc_count(inode);
- atomic_inc(&inode->i_count);
- d_instantiate(dentry, inode);
- ll_invalidate_inode_pages(dir);
- RETURN(0);
- }
+ ENTRY;
+ CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p),dir=%lu/%u(%p),target=%s\n",
+ src->i_ino, src->i_generation, src,
+ dir->i_ino, dir->i_generation, dir, name);
- if (S_ISDIR(inode->i_mode))
- return -EPERM;
+ ll_prepare_mdc_op_data(&op_data, src, dir, name, len, 0);
+ err = md_link(sbi->ll_mdc_exp, &op_data, &request);
+ ptlrpc_req_finished(request);
- if (inode->i_nlink >= EXT2_LINK_MAX)
- return -EMLINK;
+ RETURN(err);
+}
- rc = ll_mdc_link(old_dentry, dir,
- dentry->d_name.name, dentry->d_name.len);
- if (rc)
- RETURN(rc);
- inode->i_ctime = CURRENT_TIME;
- ext2_inc_count(inode);
- atomic_inc(&inode->i_count);
+static int ll_mkdir_raw(struct nameidata *nd, int mode)
+{
+ struct inode *dir = nd->dentry->d_inode;
+ const char *name = nd->last.name;
+ int len = nd->last.len;
+ struct ptlrpc_request *request = NULL;
+ struct ll_sb_info *sbi = ll_i2sbi(dir);
+ struct mdc_op_data op_data;
+ int err = -EMLINK;
+ ENTRY;
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
+ name, dir->i_ino, dir->i_generation, dir);
- return ext2_add_nondir(dentry, inode);
+ mode = (mode & (S_IRWXUGO|S_ISVTX) & ~current->fs->umask) | S_IFDIR;
+ ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
+ err = md_create(sbi->ll_mdc_exp, &op_data, NULL, 0, mode,
+ current->fsuid, current->fsgid, 0, &request);
+ ptlrpc_req_finished(request);
+ RETURN(err);
}
-static int ll_mkdir(struct inode *dir, struct dentry *dentry, int mode)
+static int ll_rmdir_raw(struct nameidata *nd)
{
- struct lookup_intent *it;
- struct inode * inode;
- int err = -EMLINK;
+ struct inode *dir = nd->dentry->d_inode;
+ const char *name = nd->last.name;
+ int len = nd->last.len;
+ struct ptlrpc_request *request = NULL;
+ struct mdc_op_data op_data;
+ int rc;
ENTRY;
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
+ name, dir->i_ino, dir->i_generation, dir);
- LL_GET_INTENT(dentry, it);
-
- if (dir->i_nlink >= EXT2_LINK_MAX)
- goto out;
+ ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, S_IFDIR);
+ rc = md_unlink(ll_i2sbi(dir)->ll_mdc_exp, &op_data, &request);
+ ptlrpc_req_finished(request);
+ RETURN(rc);
+}
- ext2_inc_count(dir);
- inode = ll_create_node(dir, dentry->d_name.name, dentry->d_name.len,
- NULL, 0, S_IFDIR | mode, 0, it, NULL);
- err = PTR_ERR(inode);
- if (IS_ERR(inode))
- goto out_dir;
+int ll_objects_destroy(struct ptlrpc_request *request, struct inode *dir)
+{
+ struct mds_body *body;
+ struct lov_mds_md *eadata;
+ struct lov_stripe_md *lsm = NULL;
+ struct obd_trans_info oti = { 0 };
+ struct obdo *oa;
+ int rc;
+ ENTRY;
- ext2_inc_count(inode);
+ /* req is swabbed so this is safe */
+ body = lustre_msg_buf(request->rq_repmsg, 0, sizeof(*body));
- err = ext2_make_empty(inode, dir);
- if (err)
- goto out_fail;
+ if (!(body->valid & OBD_MD_FLEASIZE))
+ RETURN(0);
- /* no directory data updates when intents rule */
- if (!it || !it->it_disposition) {
- err = ll_add_link(dentry, inode);
- if (err)
- goto out_fail;
+ if (body->eadatasize == 0) {
+ CERROR("OBD_MD_FLEASIZE set but eadatasize zero\n");
+ GOTO(out, rc = -EPROTO);
}
- d_instantiate(dentry, inode);
-out:
- EXIT;
- return err;
-
-out_fail:
- ext2_dec_count(inode);
- ext2_dec_count(inode);
- iput(inode);
- EXIT;
-out_dir:
- ext2_dec_count(dir);
- EXIT;
- goto out;
-}
-
-static int ll_common_unlink(struct inode *dir, struct dentry *dentry,
- struct lookup_intent *it, __u32 mode)
-{
- struct inode *inode = dentry->d_inode;
- struct ext2_dir_entry_2 * de;
- struct page * page;
- int rc = 0;
-
- if (it && it->it_disposition) {
- rc = it->it_status;
- ll_invalidate_inode_pages(dir);
- if (rc)
- GOTO(out, rc);
- GOTO(out_dec, 0);
+ /* The MDS sent back the EA because we unlinked the last reference
+ * to this file. Use this EA to unlink the objects on the OST.
+ * It's opaque so we don't swab here; we leave it to obd_unpackmd() to
+ * check it is complete and sensible. */
+ eadata = lustre_swab_repbuf(request, 1, body->eadatasize, NULL);
+ LASSERT(eadata != NULL);
+ if (eadata == NULL) {
+ CERROR("Can't unpack MDS EA data\n");
+ GOTO(out, rc = -EPROTO);
}
- de = ext2_find_entry(dir, dentry, &page);
- if (!de)
- GOTO(out, rc = -ENOENT);
- rc = ll_mdc_unlink(dir, dentry->d_inode, mode,
- dentry->d_name.name, dentry->d_name.len);
- if (rc)
+ rc = obd_unpackmd(ll_i2obdexp(dir), &lsm, eadata, body->eadatasize);
+ if (rc < 0) {
+ CERROR("obd_unpackmd: %d\n", rc);
GOTO(out, rc);
+ }
+ LASSERT(rc >= sizeof(*lsm));
+
+ oa = obdo_alloc();
+ if (oa == NULL)
+ GOTO(out_free_memmd, rc = -ENOMEM);
+
+ oa->o_id = lsm->lsm_object_id;
+ oa->o_gr = lsm->lsm_object_gr;
+ oa->o_mode = body->mode & S_IFMT;
+ oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE | OBD_MD_FLGROUP;
+
+ if (body->valid & OBD_MD_FLCOOKIE) {
+ oa->o_valid |= OBD_MD_FLCOOKIE;
+ oti.oti_logcookies =
+ lustre_msg_buf(request->rq_repmsg, 2,
+ sizeof(struct llog_cookie) *
+ lsm->lsm_stripe_count);
+ if (oti.oti_logcookies == NULL) {
+ oa->o_valid &= ~OBD_MD_FLCOOKIE;
+ body->valid &= ~OBD_MD_FLCOOKIE;
+ }
+ }
- rc = ext2_delete_entry(de, page);
+ rc = obd_destroy(ll_i2obdexp(dir), oa, lsm, &oti);
+ obdo_free(oa);
if (rc)
- GOTO(out, rc);
-
- /* AED: not sure if needed - directory lock revocation should do it
- * in the case where the client has cached it for non-intent ops.
- */
- ll_invalidate_inode_pages(dir);
-
- inode->i_ctime = dir->i_ctime;
-out_dec:
- ext2_dec_count(inode);
-out:
+ CERROR("obd destroy objid "LPX64" error %d\n",
+ lsm->lsm_object_id, rc);
+ out_free_memmd:
+ obd_free_memmd(ll_i2obdexp(dir), &lsm);
+ out:
return rc;
}
-static int ll_unlink(struct inode *dir, struct dentry *dentry)
-{
- struct lookup_intent * it;
-
- LL_GET_INTENT(dentry, it);
-
- return ll_common_unlink(dir, dentry, it, S_IFREG);
-}
-
-static int ll_rmdir(struct inode *dir, struct dentry *dentry)
+static int ll_unlink_raw(struct nameidata *nd)
{
- struct inode * inode = dentry->d_inode;
- struct lookup_intent *it;
+ struct inode *dir = nd->dentry->d_inode;
+ const char *name = nd->last.name;
+ int len = nd->last.len;
+ struct ptlrpc_request *request = NULL;
+ struct mdc_op_data op_data;
int rc;
ENTRY;
+ CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
+ name, dir->i_ino, dir->i_generation, dir);
- LL_GET_INTENT(dentry, it);
-
- if ((!it || !it->it_disposition) && !ext2_empty_dir(inode))
- RETURN(-ENOTEMPTY);
-
- rc = ll_common_unlink(dir, dentry, it, S_IFDIR);
- if (!rc) {
- inode->i_size = 0;
- ext2_dec_count(inode);
- ext2_dec_count(dir);
- }
+ ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
+ rc = md_unlink(ll_i2sbi(dir)->ll_mdc_exp, &op_data, &request);
+ if (rc)
+ GOTO(out, rc);
+ rc = ll_objects_destroy(request, dir);
+ out:
+ ptlrpc_req_finished(request);
RETURN(rc);
}
-static int ll_rename(struct inode * old_dir, struct dentry * old_dentry,
- struct inode * new_dir, struct dentry * new_dentry)
+static int ll_rename_raw(struct nameidata *oldnd, struct nameidata *newnd)
{
- struct lookup_intent *it;
- struct inode * old_inode = old_dentry->d_inode;
- struct inode * tgt_inode = new_dentry->d_inode;
- struct page * dir_page = NULL;
- struct ext2_dir_entry_2 * dir_de = NULL;
- struct ext2_dir_entry_2 * old_de;
- struct page * old_page;
+ struct inode *src = oldnd->dentry->d_inode;
+ struct inode *tgt = newnd->dentry->d_inode;
+ const char *oldname = oldnd->last.name;
+ int oldlen = oldnd->last.len;
+ const char *newname = newnd->last.name;
+ int newlen = newnd->last.len;
+ struct ptlrpc_request *request = NULL;
+ struct ll_sb_info *sbi = ll_i2sbi(src);
+ struct mdc_op_data op_data;
int err;
+ ENTRY;
+ CDEBUG(D_VFSTRACE, "VFS Op:oldname=%s,src_dir=%lu/%u(%p),newname=%s,"
+ "tgt_dir=%lu/%u(%p)\n", oldname, src->i_ino, src->i_generation,
+ src, newname, tgt->i_ino, tgt->i_generation, tgt);
- LL_GET_INTENT(new_dentry, it);
-
- if (it && it->it_disposition) {
- if (tgt_inode) {
- tgt_inode->i_ctime = CURRENT_TIME;
- tgt_inode->i_nlink--;
- }
- ll_invalidate_inode_pages(old_dir);
- ll_invalidate_inode_pages(new_dir);
- GOTO(out, err = it->it_status);
- }
-
- err = ll_mdc_rename(old_dir, new_dir, old_dentry, new_dentry);
- if (err)
- goto out;
-
- old_de = ext2_find_entry (old_dir, old_dentry, &old_page);
- if (!old_de)
- goto out;
-
- if (S_ISDIR(old_inode->i_mode)) {
- err = -EIO;
- dir_de = ext2_dotdot(old_inode, &dir_page);
- if (!dir_de)
- goto out_old;
- }
-
- if (tgt_inode) {
- struct page *new_page;
- struct ext2_dir_entry_2 *new_de;
-
- err = -ENOTEMPTY;
- if (dir_de && !ext2_empty_dir (tgt_inode))
- goto out_dir;
-
- err = -ENOENT;
- new_de = ext2_find_entry (new_dir, new_dentry, &new_page);
- if (!new_de)
- goto out_dir;
- ext2_inc_count(old_inode);
- ext2_set_link(new_dir, new_de, new_page, old_inode);
- tgt_inode->i_ctime = CURRENT_TIME;
- if (dir_de)
- tgt_inode->i_nlink--;
- ext2_dec_count(tgt_inode);
- } else {
- if (dir_de) {
- err = -EMLINK;
- if (new_dir->i_nlink >= EXT2_LINK_MAX)
- goto out_dir;
- }
- ext2_inc_count(old_inode);
- err = ll_add_link(new_dentry, old_inode);
- if (err) {
- ext2_dec_count(old_inode);
- goto out_dir;
- }
- if (dir_de)
- ext2_inc_count(new_dir);
+ ll_prepare_mdc_op_data(&op_data, src, tgt, NULL, 0, 0);
+ err = md_rename(sbi->ll_mdc_exp, &op_data,
+ oldname, oldlen, newname, newlen, &request);
+ if (!err) {
+ err = ll_objects_destroy(request, src);
}
- ext2_delete_entry (old_de, old_page);
- ext2_dec_count(old_inode);
-
- if (dir_de) {
- ext2_set_link(old_inode, dir_de, dir_page, new_dir);
- ext2_dec_count(old_dir);
- }
- return 0;
+ ptlrpc_req_finished(request);
-out_dir:
- if (dir_de) {
- kunmap(dir_page);
- page_cache_release(dir_page);
- }
-out_old:
- kunmap(old_page);
- page_cache_release(old_page);
-out:
- return err;
+ RETURN(err);
}
struct inode_operations ll_dir_inode_operations = {
- create: ll_create,
- lookup2: ll_lookup2,
- link: ll_link,
- unlink: ll_unlink,
- symlink: ll_symlink,
- mkdir: ll_mkdir,
- rmdir: ll_rmdir,
- mknod: ll_mknod,
- rename: ll_rename,
- setattr: ll_setattr
+ .link_raw = ll_link_raw,
+ .unlink_raw = ll_unlink_raw,
+ .symlink_raw = ll_symlink_raw,
+ .mkdir_raw = ll_mkdir_raw,
+ .rmdir_raw = ll_rmdir_raw,
+ .mknod_raw = ll_mknod_raw,
+ .mknod = ll_mknod,
+ .rename_raw = ll_rename_raw,
+ .setattr = ll_setattr,
+ .setattr_raw = ll_setattr_raw,
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
+ .create_it = ll_create_it,
+ .lookup_it = ll_lookup_it,
+ .revalidate_it = ll_inode_revalidate_it,
+#else
+ .lookup = ll_lookup_nd,
+ .create = ll_create_nd,
+ .getattr_it = ll_getattr,
+#endif
};